IMAGE DATA COMPRESSION DEVICE, COMPRESSION METHOD, PROGRAM, AND IMAGE DATA DECOMPRESSION DEVICE, DECOMPRESSION METHOD, AND PROGRAM

Description

Technical Field

[0001] The present invention relates to an image data compression device and a compressing method capable of processing two types (natural images such as movies etc. and CG images such as a digital map etc.) of high quality images having different characteristics in real time, and a program for performing the compression, and an image data decompression device and a decompressing method for decompressing compressed image data, and a program for decompressing compressed image data.

Background Art

[0002] It is necessary to compress data to transmit image information containing an large amount of information. Especially, since 30 through 60 frames of image data are transmitted and received for moving pictures, the data compression is required.

[0003] It is well known that image information can be natural images such as general television images, movies, etc. and CG images (CG images) represented by maps for car navigation etc. Generally, natural images contain a number of low-frequency components, and digital images contain a number of high-frequency components. In a recent mobile terminal such as a car-mounted terminal, a mobile telephone, etc., both digital images such as a map etc. and natural images such as TV and movie pictures etc. are processed. To efficiently transmit both types of image data, an effective data compressing system for both low-frequency components and high-frequency components is desired.

[0004] The first conventional technology of compressing data using a JPEG (joint photographic experts group) and MPEG (moving picture experts group) shown in FIG. 1 is known as a common conventional image data compressing system. The following patent documents 1 and 2 belong to the first conventional technology. As shown in FIG. 1, the first conventional technology divides image data 110 into blocks (normally 8*8 pixels), performs frequency conversion 210 on the block image, performs quantization 310 on a DCT coefficient, performs variable-length encoding 410 for assigning a code depending on the frequency of occurrence, and transmits the result. The frequency conversion is to convert the frequency of image data normally using a DCT (discrete cosine transfer). Since human eyes are sensitive to low-frequency components (flat portion in an image), the DCT coefficient for a low frequency is precisely quantized, and the DCT coefficient for a high frequency is roughly quantized, thereby compressing natural images at a high compression rate at which image degradation can be inconspicuous. However, there is no problem with the compression of natural images when conspicuous low-frequency components are precisely quantized, but the image degradation is conspicuous on the high-frequency components such as a line and a character in map images (CG images). Furthermore, since edge information extraction 230 is performed on a block to be compressed, there is the problem that correction and feedback cannot be performed by detecting the image degradation caused by a quantization error. Therefore, it is not appropriate for CG images.

[0005] In addition, the second conventional technology using the JPEG-LS (lossless) as shown in FIG. 2 is known as a conventional image data compressing system. The second conventional technology predicts a level value using a MED (median edge detector) prediction device (the prediction device is a type of MAP (median adaptive predictor) from a level value of a pixel to be compressed when the image data 120 is compressed as shown in FIG. 2, and directly encodes prediction error. In this second conventional technology, since the frequency of occurrence of the prediction error basically concentrates on the vicinity of 0, the prediction error in the vicinity of 0 is assigned a short code while a prediction error having a large value is assigned a long code for compression. Although the compression rate is about 1/3 , the encoding process is performed in a pixel unit regardless of high or low frequencies. Accordingly, no image degradation occurs even on CG images.

[0006] However, it is difficult to predict an interlace image and an image having low correlation between lines, and prediction error normally increases on these images. As a result, a variable-length code having a low frequency of occurrence (long code length) is assigned, and it is necessary to perform calculating process during encoding. Thus, each process becomes heavy and is not appropriate for real-time compressing process.

Patent Document 1: Japanese Published Patent Application No. 2001-061149

Patent Document 2: Japanese Published Patent Application No. 2000-069479

[0007] Torsten Seemann, Peter Tischer in "Generalized Locally Adaptive DPCM", 25 January 2000, pages 1 - 15, URL: http:www.csse.monash.edu.au/ {torsten/pubs/TR301.ps.gz>, shows an image data compression device according to the preamble of claim 1, wherein an adaptive predictor is one for which coefficients are varied for each prediction event based on some assessment of the local neighborhood.

Disclosure of the Invention

[0008] The prediction error is conspicuous on CG images when transfer or coding is performed by DCT such as the JPEG, MPEG, etc. while the prediction error on an interlace image and a landscape image is troublesome. Therefore, a long code is frequently used, and reduces compression efficiency.

[0009] Thus, a car-mounted image data compression device for processing both natural images and CG images has to compress CG images with high quality, and maintain the compression efficiency for interlace images and landscape images in movies and natural images.

[0010] To solve the above-mentioned problems, the present invention aims at providing an image data compression device

[0011] and compressing method for predicting the level value of a pixel to be predicted from the level value of a pixel around the pixel to be predicted, and quantizing the error of a prediction result to compress data, and a program for performing the compression, and a image data decompression device and decompressing method for decompressing compressed image data, and a program for performing the decompression.

[0012] The image data compression device according to the present invention is defined in claim 1. A prediction error can be reduced for each image format, and the compression efficiency can be enhanced. In addition, since only a prediction value is to be switched depending on the image format, the program for implementation and the configuration of a circuit can be simplified.

[0013] In addition, the image data compressing method according to the present invention is defined in claim 4 In each of the prediction value determination module and the prediction value calculation module, parallel processing is performed, and when the calculation of each candidate for a prediction value is completed, the prediction value to be used can be determined. Therefore, data can be compressed at a high speed by the parallel processing.

[0014] The program according to the present invention is defined in claim 6 parallel processing is performed in each step of the prediction value determining and prediction value calculating steps, and it can be determined which prediction value is to be used when each candidate for a prediction value is completed. Therefore, data can be compressed at a high speed by the parallel processing.

[0015] In addition, since the image data decompression device and the image data decompressing method according to the present invention decompress image data compressed by the image data compression device in the reverse operation, the prediction error can be reduced for each image format as with the image data compression device, thereby enhancing the compression efficiency. Additionally, only by switching the prediction value depending on the image format, a program for implementation and the configuration of a circuit can be simplified.

[0016] Since the program for decompressing image data according to the present invention decompresses image data compressed by the above-mentioned program for compressing image data in the reverse operation, as with the above-mentioned program for compressing image data, parallel processing is performed in each step of the prediction value determining and prediction value calculating steps, and it can be determined which prediction value is to be used when each candidate for a prediction value is completed. Therefore, data can be compressed at a high speed by the parallel processing.

Brief Description of the Drawings

[0017] 

FIG. 1 shows the first conventional technology for compressing data by the JPEG and MPEG;

FIG. 2 shows the second conventional technology for compressing data by the JPEG-LS;

FIG. 3 is a block diagram of the configuration of the image data compression device according to an embodiment of the present invention;

FIG. 4 shows the arrangement of a pixel to be compressed and peripheral pixels in the image data compression device according to an embodiment of the present invention;

FIG. 5 shows a quantization table in the quantizer according to an embodiment of the present invention;

FIG. 6 shows the inverse quantization table in the prediction value determination module of the image data compression device according to an embodiment of the present invention;

FIG. 7 shows the intermediate value - image format correspondence table in the prediction value determination module of the image data compression device according to an embodiment of the present invention;

FIG. 8 shows a correspondence table indicating the correspondence between the control signal of the image data compression device and the candidate for candidates 1 through k for a prediction value according to an embodiment of the present invention;

FIG. 9A is a flowchart for explanation of the operation of the image data compression (encoding) device according to an embodiment of the present invention;

FIG. 9B is a flowchart for explanation of the process of the "local decoder" in step S008 in FIG. 9A;

FIG. 10 shows a pixel to be compressed in the image data compression device and the pixel level value at a time point of a line buffer for prediction according to an embodiment of the present invention;

FIG. 11 shows a quantization result of the prediction device for an interlace image in the image data compression device according to an embodiment of the present invention;

FIG. 12 shows a quantization result of the prediction device for a progressive image in the image data compression device according to an embodiment of the present invention;

FIG. 13 is a block diagram of the configuration of the system of the image data compression device according to an embodiment of the present invention;

FIG. 14 is a block diagram of the configuration of the image data decompression device according to an embodiment of the present invention;

FIG. 15 shows an inverse quantization table in the inverse-quantizer according to an embodiment of the present invention;

FIG. 16 shows the arrangement of the pixel to be decompressed and the peripheral pixels in the image data decompression device according to an embodiment of the present invention;

FIG. 17 shows an inverse quantization table in the prediction value determination module of the image data decompression device according to an embodiment of the present invention;

FIG. 18 shows an intermediate value - image format correspondence table in the prediction value determination module of the image data decompression device according to an embodiment of the present invention;

FIG. 19 shows a correspondence table indicating the correspondence between the control signal and the candidates 1 through k for a prediction value of the image data decompression device according to an embodiment of the present invention;

FIG. 20A is a flowchart for explanation of the operation of the image data decompression (decoding) device according to an embodiment of the present invention;

FIG. 20B is a flowchart for explanation of the process of the "local decoder" in step S027 in FIG. 20A;

FIG. 21 shows the pixel level value at a time point of a line buffer for prediction in the image data compression device and according to an embodiment of the present invention;

FIG. 22 shows a quantization result of the prediction device for an interlace image in the image data decompression device according to an embodiment of the present invention;

FIG. 23 shows a quantization result of the prediction device for a progressive image in the image data decompression device according to an embodiment of the present invention;

FIG. 24 is a block diagram of the configuration of the system of the image data decompression device according to an embodiment of the present invention;

Best Mode for Carrying Out the Invention

[0018] The embodiments of the present invention are described below with reference to the attached drawings.

[0019] FIG. 3 is a block diagram of the configuration of the image data compression (encoding) device according to an embodiment of the present invention. In FIG. 3, image data 001 indicates a group of pixels to be compressed, and is extracted by an image processing device (not shown in the attached drawings) for digitally processing images. Then, a pixel X (002) to be compressed indicates a pixel to be compressed in the current step. FIG. 4 shows the arrangement of a pixel X to be compressed and peripheral pixels in the image format of an interlace image or a progressive image. In FIG. 4, the lines are indicated by broken lines. A quantizer 003 quantizes a "prediction error" as a difference value between the pixel X (002) to be compressed and a prediction value X' (006).

[0020] FIG. 5 shows an example of a quantization table indicating the correspondence between the prediction error (X - X') and the quantization value and the quantization number according to an embodiment of the present invention. The quantization table shown in FIG. 5 receives a prediction error (X - X'), and outputs a prediction error quantization value (also referred to simply as a quantization value) and a quantization number. It is desired that the widths of the quantizing steps are narrower for smaller absolute values of prediction error, and wider for larger absolute values of prediction error for the following reason. Generally, when a prediction error is small, it refers to a flat image which can be easily predicted. On the other hand, when a prediction error is large, it refers to the vicinity of an edge which cannot be easily predicted. If there is an error in a flat portion, it can be easily detected by human eyes. Therefore, when a prediction error is small, it is necessary to narrow the step width. It is desired that the minimum step width is 1 through 4 depending on the necessity of the adaptive quantization. On the other hand, it is desired that the step width is 24 through 32 when prediction error is large because the frequency of occurrence is low and it is necessary to represent one quantizing step for a large number of prediction errors to enhance the compression efficiency. In addition, when a complicated pattern having a number of edges is compressed, there is the possibility that a predetermined compression rate cannot be attained. Therefore, when a predetermined compression rate can be exceeded by a plurality of quantization tables having different quantizing step widths, adaptive quantization of selecting a totally rough quantization table in step width can be performed. Furthermore, it is necessary that the correspondence between the quantization value and the quantization number shown in FIG. 5 is consistent between the compressing side and the decompressing side. Therefore, the decompressing side is provided with an inverse quantization table (FIG. 15) showing the correspondence between the quantization value and the quantization number with respect to FIG. 5. It is described later.

[0021] A variable-length encoder 004 receives a quantization number obtained from the quantization table shown in FIG. 5 and outputs a variable-length code. A compressed code buffer 005 temporarily stores the variable-length code output from the variable-length encoder 004, that is, the compressed code.

[0022] The prediction value X' (006) is one of the candidates for a prediction value selected by a multiplexer (MPX) 015 from among the candidates 1 through k (k=6 in the description of the present embodiment) for a prediction value calculated by a prediction value candidate calculation module 011. A peripheral pixel A (007) is a pixel left to the pixel X (002) to be compressed, that is, a previous pixel processed in the same line as shown in FIG. 4. A line buffer 008 for prediction holds about 1 line of quantization results for prediction, and is configured by, for example, a shift register. As shown in FIG. 4, a peripheral pixel C (009) is a pixel above the pixel X to be compressed, that is, a pixel in the same column in the line processed in the preceding line. A peripheral pixel B (010) is a pixel upper left to the pixel X to be compressed, that is, a previous pixel processed to the pixel C (009) in the previous line processed. In the original image, the peripheral pixel C and the peripheral pixel B are pixels in the second line above the pixel X to be compressed in the interlace image and in the line immediately above in the progressive image.

[0023] The prediction value X' (006) is one of the candidates for a prediction value calculated by the prediction value candidate calculation module 011 on the basis of the peripheral pixels A (007), C (009), and B (010), but the calculation of a candidate for a prediction value when the image format is a progressive image is first described below. As shown in FIG. 4, in the case of a progressive image, using the pixel level values of the two pixels C and B in the previous line processed and the prediction level value of the previous pixel processed in the same line, the candidate for a prediction value for the prediction value X' (006) is A, C, (A+C-B) with the correlation among three directions taken into account.

[0024] Described next is the calculation of a candidate for a prediction value when the image format is an interlace image. As shown in FIG. 4, in the case of an interlace image, an even line and an odd line are alternately transmitted. Therefore, when processing is performed in a line unit without frame memory, pixels processed in the second line above, which have weaker correlations, are used. Therefore, without using the pixel level values of the pixels C and B in the second previous line processed, the pixels C' (pixel between the pixel X and the pixel C) and B' (pixel between the pixel A and the pixel B) are considered. To process the pixels as in the case of the above-mentioned progressive image, the pixel level value of the pixel C' is defined as an interpolation value " (A+C) /2" between the pixels A and C, and the level value of the pixel B' is calculated using the interpolation value " (2A+C-B) /2" between the pixels A and B. Furthermore, the candidate for a prediction value for the prediction value X' (006) is defined as A, (A+C)/2, (2A+C-B) /2 using the pixel level value of the pixel A as the previous pixel processed in the same line in addition to the pixel C' and the pixel B', that is, with the correlation among the three directions taken into account.

[0025] Thus, the prediction value candidate calculation module 011 calculates the following candidates 1 through k for a prediction value on the basis of the peripheral pixels A (007), C (009), and B (010), and transmits the calculation results to the multiplexer (MPX) 015.

candidate for a prediction value 1: (A+C)/2

candidate for a prediction value 2: A

candidate for a prediction value 3: (2A+C-B)/2

candidate for a prediction value 4: C

candidate for a prediction value 5: A

candidate for a prediction value 6: A+C-B

[0026] In the calculation above, the prediction value candidate calculation module 011 can determine a prediction value for an interlace image as an intermediate value of three candidates for a prediction value "A", "(m*A+n*C)/(m+n)", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n)" using coefficients m and n that are values (positive integers including 0) other than 0, and a sum of m and n is a value of i-th power of 2 so that the process in the shift operation can be performed. To obtain the above-mentioned prediction value, the prediction value candidate calculation module 011 calculates the three candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n)" as the candidates for a prediction value 7 through 9 in addition to the above-mentioned candidates for a prediction value 1 through 6. As a result, the value of k is 9. That is, to an interlace image with the coefficients m and n taken into account, one image format is added in addition to the interlace image or the progressive image without the consideration of the above-mentioned coefficients. Therefore, the value of k is 9. Furthermore, if the number of image format increases, the number of k is three times the number of the image formats. As the coefficients m and n, for example, as an interlace-oriented prediction value using m=1 and n=2, a prediction value that is strong against the fluctuation of the pixel level value in the vertical direction can be calculated using "A", "(A+3*C)/4", and "(4*A+3*C-3*B)/4". That is, for the image having a strong correlation in the vertical direction is processed using the value of n larger than the value of m. The values of the above-mentioned coefficients in the case of a progressive image are m=0 and n=1 so that a prediction value similar to the value obtained for the conventional MAP prediction device can be acquired.

[0027] The image format 012 is presented from the image processing device (not shown in the attached drawings) to the image data compression device. However, an image data format is generally represented as a signal indicating in which direction the image has a strong pixel correlation, the horizontal direction or the vertical direction. That is, the above-mentioned interlace image has a strong pixel correlation in the horizontal direction, and the above-mentioned progressive image has a strong pixel correlation in the vertical direction (the pixel correlation in the horizontal direction is not necessarily stronger than the pixel correlation in the vertical direction). The same holds true with the case in which the image format is a sampling format of the image data according to the signal.

[0028] The prediction value determination module 013 determines which in the candidates 1 through k for a prediction value is set as a prediction value on the basis of two correspondence tables shown in FIGS. 6 and 7, and transmits it as a control signal 014 to the multiplexer (MPX) 015. FIG. 6 is an intermediate value correspondence table for comparing the pixel level values of the peripheral pixels A, B, and C, calculating the intermediate value of the peripheral pixels A, B, and C, and narrowing the candidates for a prediction value into the candidates (1) through (3) for a prediction value. The correspondence between the candidates (1) through (3) for a prediction value and the intermediate value shown in FIG. 6 has to be consistent between the compressing side and the decompressing side. Therefore, with respect to FIG. 6, the decompressing side is provided with the intermediate value correspondence table (FIG. 17) indicating the correspondence between the candidates (1) through (3) for a prediction value and an intermediate value. It is described later.

[0029] FIG. 7 is an intermediate value - image format correspondence table used in transmitting to the multiplexer (MPX) 015 the candidates (1) through (3) for a prediction value narrowed in FIG. 6 and the control signal 014 indicating which in the candidates 1 through k for a prediction value is finally to be used as a prediction value depending on whether the image format 012 is an image (interlace image) having a strong pixel correlation in the horizontal direction or an image (progressive image) having a strong pixel correlation in the vertical direction. What is output as the control signal 014 is a 3-bit (000 - 101) signal as shown in the correspondence table in FIG. 7. In addition, as described above, when the calculation output of the three candidates for a prediction value "A", "(m*A+n*C)/(m+n)", and "A+(m*A+n*C)/(m+n)-(m*A+n*B)/(m+n)" about an interlace image with the coefficients m and n taken into account is added, the candidates for a prediction value "A", "(m*A+n*C)/(m+n)", and "A+(m*A+n*C)/(m+n)-(m*A+n*B)/(m+n)" identified by a control signal (represented by 4 bits in this case) are added to the image column having a strong pixel correlation in the horizontal direction with respect to the candidates (1) through (3) for a prediction value. In addition, the correspondence between the intermediate value an the image format shown in FIG. 7 is to be consistent between the compressing side and the decompressing side. Therefore, with respect to FIG. 7, the decompressing side is provided with an intermediate value - image format correspondence table (FIG. 18) indicating the correspondence between the intermediate value and the image format. It is described later.

[0030] The multiplexer (MPX) 015 determines which of the candidates 1 through k for a prediction value is selected as a prediction value according to the control signal 014. FIG. 8 is a correspondence table showing the correspondence between the control signal 014 and the above-mentioned candidates for a prediction value 1 through 6. That is, a control signal (000) corresponds to a candidate 1 for a prediction value, a control signal (001) corresponds to a candidate 2 for a prediction value, a control signal (010) corresponds to a candidate 3 for a prediction value, a control signal (011) corresponds to a candidate 4 for a prediction value, a control signal (100) corresponds to a candidate 5 for a prediction value, and a control signal (101) corresponds to the candidate 6 for a prediction value. As described above, when the calculation output of the three candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n) " for an interlace image with the coefficients m and n taken into account are added from the prediction value candidate calculation module 011, the candidates for a prediction value "A", "(m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n)" are added as the candidates for a prediction value to be used in the table shown in FIG. 8. In addition, a control signal (a 4-bit signal in this case) having a bit width for identification of the three candidates for a prediction value with the control signal added is required. The correspondence between the control signal and the candidate for a prediction value shown in FIG. 8 is to be consistent between the compressing side and the decompressing side. With respect to FIG. 8, a correspondence table (FIG. 19) showing the correspondence between the control signal and the candidate for a prediction value is provided for the decompressing side. It is described later. Although 3-bit control signals (000) through (101) are associated as the control signal 014 according to the present embodiment, the control signal not in this format can also be used so far as the candidates 1 through 6 for a prediction value can be associated.

[0031] In addition, a correspondence table indicating according to the signal about a sampling format of image data which prediction value is to be used is stored in the prediction value determination module 013 regardless of the interlace image or the progressive image, and a prediction value can be determined on the basis of the correspondence table.

[0032] Although not shown in the attached drawings, a control signal indicating the sampling format of the image data is transmitted from the prediction value determination module 013 to the prediction value candidate calculation module 011. Based n the signal, the prediction value candidate calculation module 011 can determine the value of the above-mentioned coefficients m and n.

[0033] FIG. 9A is a flowchart for explanation of the operation of the image data compression (coding) device according to an embodiment of the present invention. FIG. 9B is a flowchart for explanation of the process of the "local decoder" in step S008 shown in FIG. 9A. In FIGS. 9A and 9B, S is short for step.

[0034] Before describing the steps shown in FIG. 9A, it is to be noted that the process enters the loop until all image data is processed in the present embodiment.
S001: The pixel X (002) to be compressed is acquired from the image data 001.
S002: The continuity evaluation result feedback module 013 compares the pixel level values among the peripheral pixels A (007), C (009), and B (010). Based on the comparison result and the image format 012, and with reference to the correspondence tables shown in FIGS. 6 and 7, the control signal 014 is transmitted to the multiplexer (MPX) 015.
S003: The prediction value candidate calculation module 011 calculates the candidates 1 through 6 for a prediction value on the basis of the peripheral pixels A (007), C (009), and B (010). In this case, a prediction value is calculated by an integer operation and a shift operation. Then, during the integer operation of (2A+C-B+1) / 2 and (A+C+1) / 2, the digits to the right of the decimal point are discarded. During the real number operation, 1 is added, and no digits are discarded and the value of (2A+C-B)/2 and (A+C)/2 are used as is.
S004: The MPX 015 determines a prediction value according to the candidates 1 through 6 for a prediction value and the control signal 014.
S005: The prediction value 006 is subtracted from the pixel level value of the pixel X 002 to be compressed, and the prediction error is calculated.
S006: The prediction error is quantized by the quantizer 003, and a quantization value and a quantization number are acquired.
S007: A variable-length code is generated on the basis of the quantization number using the variable-length encoder 004. The code can be a Golomb code, or an arithmetic code.
S008: To compress the next pixel, the peripheral pixels A, B, and C, and the line buffer 008 for prediction are updated (local decoder).
S009: The peripheral pixel C (009) is substituted for the peripheral pixel B (010) of the next pixel to be compressed.
S010: The peripheral pixel C (009) of the next pixel to be compressed is acquired from the line buffer 008 for prediction.
S011: The peripheral pixel A (007) is substituted for the line buffer 008 for prediction.
S012: The quantization value and the prediction value are added up, and the sum is substituted for the peripheral pixel A (007) of the next pixel to be compressed.

[0035] In addition, for example, the line buffer 008 for prediction and the pixel X (002) to be compressed refer to the values as shown in FIG. 10, quantization results of a prediction device for an interlace image and a prediction device for a progressive image are represented as the tables shown in FIGS. 11 and 12. When the line buffer 008 for prediction and the pixel X (002) to be compressed have the pixel values as shown in FIG. 10, the tables shown in FIGS. 11 and 12 can be easily derived from the operation of the image data compression (coding) device according to the embodiments of the present invention shown in FIGS. 9A and 9B. In this case, the pixel C (009) after the previous compression is the pixel B (010) after the current compression, and the previous "prediction value + prediction error quantization value" is the current pixel A (007).

[0036] FIG. 13 is a block diagram of the configuration of the system of the image data compression device according to the embodiment of the present invention, and is realized by a block represented as the function of the contents described above. In FIG. 13, the image data compression device according to the embodiment of the present invention inputs the image data 102 to be compressed to the compressing unit 110, the compressing unit 110 first reads the input image data 102 by the read unit 111 for each line to extract a pixel to be compressed. The predicting unit 112 calculates on the extracted pixel to be compressed a candidate for a prediction value by a prediction value candidate calculation unit 115 corresponding to the prediction value candidate calculation module oil shown in FIG. 3 with reference to an intermediate value correspondence table 113 shown in FIG. 6 and an intermediate value - image format correspondence table 114 shown in FIG. 7.

[0037] Furthermore, compressed data 106 are obtained by determining with a prediction value determination unit 116 corresponding to the MPX 015 - a prediction value among the prediction value candidates calculated on the basis of an image format 104, obtaining a prediction error from the difference between the determined prediction value and the pixel to be compressed, inputting this prediction error into a prediction value candidate calculation unit 115 corresponding to the quantizer 003, inputting at the quantizing unit 117 the quantization number by referring to the quantization table 118 as shown in FIG. 5 along with obtaining a prediction error quantization value and a quantization number from the prediction error, obtaining a variable-length code (compressed code) from the coding unit 119 corresponding to the variable-length encoder 004, accumulating this as the output of the compressing unit 110.

[0038] Thus, according to the image data compression device of the present invention, the prediction value can be appropriately used depending on the image format, and the prediction error can be reduced on each image format, thereby enhancing the compression efficiency. Since only a prediction value is to be switched depending on the image format, a program and a circuit configuration for implementation can be simplified.

[0039] In the descriptions above, an intermediate value among A, B, and C is obtained and a candidate for a prediction value to be selected is determined on the basis of an intermediate value correspondence table, but an intermediate value of candidates for a prediction value can also be obtained after calculating the candidates for a prediction value.

[0040] Also in the description above, the prediction value for an interlace image is (A+C)/2, A and (2A+C-B)/2. However, the prediction value candidate calculation module 011 calculates "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n) " using the coefficients m and n from the pixel level values A, B, and C, substitutes a value using coefficients m and n that are values (positive integers including 0) other than 0, and a sum of m and n is a value of i-th power of 2 so that the process in the shift operation can be performed, and can calculate a prediction value strong against the fluctuation of the pixel level value in the vertical direction using "A", "(A+3*C)/4", and "(4*A+3*C-3*B)/4" as the prediction value for an interlace image.

[0041] In addition, regardless of an interlace image or an progressive image as an image format, the prediction value can be used appropriately according to a signal indicating a sampling format of image data to attain similar effect.

[0042] For example, the sampling format of the image data stored on a common DVD is normally a 4:2:2 format. In this format, the image size of the Y component (brightness component) is 720 × 480, the horizontal image size of the C component (color component) is a half of the Y component. Therefore, since the C component refers to image data having a strong correlation in the vertical direction, problems can be solved by making the value of n larger than the value of m. That is, by increasing the gravity to the peripheral pixel having a strong correlation in the vertical direction (increasing the value of n), the degradation of the prediction error can avoided. For example, in the interlace image having the image data of a sampling format of 4:2:2 and the image size of 720 × 480, the image data having a strong correlation in the vertical direction is processed using m=1 and n=3 for the C component. For the Y component, image data having a strong correlation in the horizontal direction is processed using m=1 and n=1 in the case of the image size 720 × 480. Relating to the coefficients m and n when the image data has a strong correlation in the horizontal direction, the value of m is assigned the same value of n. If the degradation of the prediction error still advances, the value of m is further increased. Furthermore, when the sampling format of the image data as another example is a 4:4:4 format and the image size is 800 × 800 as a progressive image, m=0 and n=1.

[0043] Furthermore, since only field memory (buffer of 1 line) is used, and no frame memory is required, the implementation cost can be reduced.

[0044] In the description above, the prediction value candidate calculation module 011 calculates k (6 in the description above) candidates for a prediction value, and selects one prediction value by the multiplexer (MPX) 015 according to a control signal narrowed with reference to FIGS. 6 and 7. In the process order, a candidate for a prediction value can be narrowed with reference to FIGS. 6 and 7, and a prediction value can be calculated with the same timing, and the process speed can be improved. That is, in each of the prediction value determination module and the prediction value calculation module, parallel processing is performed and each candidate for a prediction value is calculated, each candidate for a prediction value is calculated, and then a prediction value to be used can be determined, thereby performing data compression at a high speed by the parallel processing. In the prediction value determination module and the prediction value calculation module, an arithmetic process is performed using a table. Therefore, program processing can be performed at a high speed.

[0045] On the other hand, as the implementation when the configuration scale is reduced without overweighting the process speed, the intermediate value correspondence table shown in FIG. 6 is removed from the configuration, three prediction values are first calculated depending on the image format with reference to FIG. 7, and a intermediate value is directly obtained from the three prediction values. Furthermore, another pattern can be calculating and determining by which equation of a candidate for a prediction value using the intermediate value correspondence table shown in FIG. 6, and performing a calculating process by the determined equation only, thereby deleting the configuration of selecting one prediction value from among a plurality of calculated prediction values.

[0046] The image data compression device according to the present invention processes image data passing like a stream in the order from upward to downward on the screen in a frame unit for a progressive image and in a field unit for an interlace image. The compressed data after the compression coding is put in a packet and transferred in accordance with the transmission rules of the transmission line connected to the present device.

[0047] FIG. 14 is a block diagram showing the configuration of the image data decompression (decoding) device according to the embodiment of the present invention. In FIG. 14, a compressed code 021 refers to a compressed code of image data variable-length encoded as the output of the image data compression (coding) device according to an embodiment of the present invention shown in FIG. 3. An inverse encoder 022 receives a compressed code of image data, and outputs a quantization number corresponding to the code. A inverse-quantizer 023 receives the quantization number and outputs a prediction error quantization value X (024).

[0048] FIG. 15 is an inverse quantization table showing the correspondence of the prediction error quantization value to the quantization number provided for the inverse-quantizer 023. The inverse quantization table shown in FIG. 15 receives the quantization number and outputs the prediction error quantization value. The correspondence between the prediction error quantization value and the quantization number shown in FIG. 15 has to be consistent between the decompressing side and the compressing side. Therefore, the compressing side is provided with a quantization table (FIG.5) showing the correspondence between the prediction error quantization value and the quantization number corresponding to FIG. 15.

[0049] FIG. 16 shows the arrangement of the pixel X to be decompressed and the peripheral pixel when an image format 032 is, for example, an interlace image or a progressive image. In FIG. 5, the lines are shown as broken lines.

[0050] The prediction error quantization value X (024) is added to a prediction value X' (026) to obtain decompressed image data 025, and can be the peripheral pixel A (027) as the next pixel to be decompressed.

[0051] A peripheral pixel A (027) is a pixel left to the pixel to be decompressed as shown in FIG. 6. A line buffer for prediction 028 holds about one line of quantization result for prediction, and is configured by, for example, a shift register. As shown in FIG. 16, a peripheral pixel C (029) is a pixel above the pixel to be decompressed (in the preceding line), and a peripheral pixel B (030) is a pixel upper left to the pixel to be decompressed (in the preceding line).

[0052] The prediction value X' (026) is one of the candidates for a prediction value calculated by the prediction value candidate calculation module 031 on the basis of the peripheral pixels A (027), C (029), and B (030), but the calculation of a candidate for a prediction value when the image format is a progressive image is first described below. As shown in FIG. 16, in the case of a progressive image, using the pixel level values of the two pixels C and B in the previous line processed and the prediction level value of the previous pixel processed in the same line, the candidate for a prediction value for the prediction value X' (026) is A, C, (A+C-B) with the correlation among three directions taken into account.

[0053] Described next is the calculation of a candidate for a prediction value when the image format is an interlace image. As shown in FIG. 16, in the case of an interlace image, an even line and an odd line are alternately transmitted. Therefore, when processing is performed in a line unit without frame memory, pixels processed in the second line above, which have weaker correlations, are used. Therefore, without using the pixel level values of the pixels C and B in the second previous line processed, the pixel C' between the pixel C and the pixel X, and the pixel B' between the pixel B and the pixel A are assumed, and the process as with the above-mentioned progressive image is to be performed using an interpolation value "(A+C)/2" between the pixels A and C and the interpolation value "(2A+C-B)/2" between the pixels A and B, and furthermore, the candidate for a prediction value for the prediction value X' (006) is defined as A, (A+C)/2, (2A+C-B)/2 using the pixel level value of the pixel A as the previous pixel processed in the same line in addition to the pixel C' and the pixel B', that is, with the correlation among the three directions taken into account.

[0054] Thus, the prediction value candidate calculation module 031 calculates the following candidates 1 through k (k=6 in the description of the present embodiment) for a prediction value on the basis of the peripheral pixels A (027), C (029), and B (030), and transmits the calculation results to the multiplexer (MPX) 035.

candidate for a prediction value 1: (A+C)/2

candidate for a prediction value 2: A

candidate for a prediction value 3: (2A+C-B)/2

candidate for a prediction value 4: C

candidate for a prediction value 5: A

candidate for a prediction value 6: A+C-B

[0055] In the calculation above, the prediction value candidate calculation module 031 can determine a prediction value for an interlace image as an intermediate value of three candidates for a prediction value "A", "(m*A+n*C)/(m+n)", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n)" using coefficients m and n that are values (positive integers including 0) other than 0, and a sum of m and n is a value of i-th power of 2 so that the process in the shift operation can be performed. To obtain the above-mentioned prediction value, the prediction value candidate calculation module 031 calculates the three candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n)" as the candidates for a prediction value 7 through 9 in addition to the above-mentioned candidates for a prediction value 1 through 6. As a result, the value of k is 9. That is, to an interlace image with the coefficients m and n taken into account, one image format is added in addition to the interlace image or the progressive image without the consideration of the above-mentioned coefficients. Therefore, the value of k is 9. Furthermore, if the number of image format increases, the number of k is three times the number of the image formats.

[0056] As the coefficients m and n, for example, as an interlace-oriented prediction value using m=1 and n=2, a prediction value that is strong against the fluctuation of the pixel level value in the vertical direction can be calculated using "A", "(A+3*C)/4", and "(4*A+3*C-3*B)/4". That is, for the image having a strong correlation in the vertical direction is processed using the value of n larger than the value of m. The values of the above-mentioned coefficients in the case of a progressive image are m=0 and n=1 so that a prediction value similar to the value obtained for the conventional MAP prediction device can be acquired.

[0057] The image format 032 is presented from the image processing device (not shown in the attached drawings) to the image data compression device. However, an image data format is generally represented as a signal indicating in which direction the image has a strong pixel correlation, the horizontal direction or the vertical direction. That is, the above-mentioned interlace image has a strong pixel correlation in the horizontal direction, and the above-mentioned progressive image has a strong pixel correlation in the vertical direction (the pixel correlation in the horizontal direction is not necessarily stronger than the pixel correlation in the vertical direction). The same holds true with the case in which the image format is a sampling format of the image data according to the signal.

[0058] The prediction value determination module 013 determines which in the candidates 1 through k for a prediction value is set as a prediction value on the basis of two correspondence tables shown in FIGS. 6 and 7, and transmits it as a control signal 014 to the multiplexer (MPX) 015. FIG. 6 is an intermediate value correspondence table for comparing the pixel level values of the peripheral pixels A, B, and C, calculating the intermediate value of the peripheral pixels A, B, and C, and narrowing the candidates for a prediction value into the candidates (1) through (3) for a prediction value. Corresponding to FIG. 17, the intermediate value correspondence table (6) showing the correspondence between the candidates for a prediction value (1) through (3) and the intermediate value is provided for the compressing side.

[0059] FIG. 18 is an intermediate value - image format correspondence table used in transmitting to the multiplexer (MPX) 015 the candidates (1) through (3) for a prediction value narrowed in FIG. 17 and the control signal 034 indicating which in the candidates 1 through k for a prediction value is finally to be used as a prediction value depending on whether the image format 032 is an interlace image or a progressive image. What is output as the control signal 034 is a 3-bit (000 - 101) signal as shown in the correspondence table in FIG. 7. In addition, as described above, when the calculation output of the three candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n) " about an interlace image with the coefficients m and n taken into account is added, the candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n) " identified by a control signal (represented by 4 bits in this case) are added to the image column having a strong pixel correlation in the horizontal direction with respect to the candidates (1) through (3) for a prediction value. Therefore, with respect to FIG. 18, the decompressing side is provided with an intermediate value - image format correspondence table (FIG. 18) indicating the correspondence between the intermediate value and the image format.

[0060] The multiplexer (MPX) 035 determines which of the candidates 1 through k for a prediction value is selected as a prediction value according to the control signal 034. FIG. 19 is a correspondence table showing the correspondence between the control signal 034 and the above-mentioned candidates for a prediction value 1 through 6. That is, a control signal (000) corresponds to a candidate 1 for a prediction value, a control signal (001) corresponds to a candidate 2 for a prediction value, a control signal (010) corresponds to a candidate 3 for a prediction value, a control signal (011) corresponds to a candidate 4 for a prediction value, a control signal (100) corresponds to a candidate 5 for a prediction value, and a control signal (101) corresponds to the candidate 6 for a prediction value. As described above, when the calculation output of the three candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n)" for an interlace image with the coefficients m and n taken into account are added from the prediction value candidate calculation module 031, the candidates for a prediction value "A", " (m*A+n*C) / (m+n) ", and "A+(m*A+n*C) / (m+n) - (m*A+n*B) / (m+n) " are added as the candidates for a prediction value to be used in the table shown in FIG. 19. In addition, a control signal (a 4-bit signal in this case) having a bit width for identification of the three candidates for a prediction value with the control signal added is required. The correspondence between the control signal and the candidate for a prediction value shown in FIG. 19 is to be consistent between the compressing side and the decompressing side. With respect to FIG. 19, a correspondence table (FIG. 19) showing the correspondence between the control signal and the candidate for a prediction value is provided for the compressing side. Although 3-bit control signals (000) through (101) are associated as the control signal 034 according to the present embodiment, the control signal not in this format can also be used so far as the candidates 1 through 6 for a prediction value can be associated.

[0061] In addition, a correspondence table indicating according to the signal about a sampling format of image data which prediction value is to be used is stored in the prediction value determination module 033 regardless of the interlace image or the progressive image, and a prediction value can be determined on the basis of the correspondence table.

[0062] Although not shown in the attached drawings, a control signal indicating the sampling format of the image data is transmitted from the prediction value determination module 033 to the prediction value candidate calculation module 031. Based n the signal, the prediction value candidate calculation module 031 can determine the value of the above-mentioned coefficients m and n.

[0063] FIG. 20A is a flowchart for explanation of the operation of the image data compression (coding) device according to an embodiment of the present invention. FIG. 20B is a flowchart for explanation of the process of the "local decoder" in step S027 shown in FIG. 20A. In FIGS. 20A and 20B, S is short for step.

[0064] Before describing the steps shown in FIG. 20A, it is to be noted that the process enters the loop until all image data is processed in the present embodiment.
S021: Compressed code 021 is input to the inverse encoder 022, and a quantization number is output.
S022: The quantization number is input to the inverse-quantizer 023, and a quantization number is output.
S023: The continuity evaluation result feedback module 033 compares the pixel level values among the peripheral pixels A (027), C (029), and B (030). Based on the comparison result and the image format 032, and with reference to the correspondence tables shown in FIGS. 17 and 18, the control signal 034 is transmitted to the multiplexer (MPX) 035.
S024: The prediction value candidate calculation module 031 calculates the candidates 1 through 6 for a prediction value on the basis of the peripheral pixels A (027), C (029), and B (030). In this case, a prediction value is calculated by an integer operation and a shift operation. Then, during the integer operation of (2A+C-B+1)/2 and (A+C+1)/2, the digits to the right of the decimal point are discarded. During the real number operation, 1 is added, and no digits are discarded and the value of (2A+C-B)/2 and (A+C)/2 are used as is.
S025: The MPX 035 determines a prediction value according to the candidates 1 through 6 for a prediction value and the control signal 034.
S026: The quantization value of the prediction error acquired in S022 is added to the prediction value acquired in S024 to calculate a pixel to be decompressed (decompressed image data 025).
S027: To decompress the next pixel, the peripheral pixels A, B, and C, and the line buffer 028 for prediction are updated (local decoder).
S028: The peripheral pixel C (029) is substituted for the peripheral pixel B (030) of the next pixel to be decompressed.
S029: The peripheral pixel C (029) of the next pixel to be decompressed is acquired from the line buffer 028 for prediction.
S030: The peripheral pixel A (027) is substituted for the line buffer 028 for prediction.
S031: The quantization value and the prediction value are added up, and the sum is substituted for the peripheral pixel A (027) of the next pixel to be decompressed.

[0065] In addition, for example, the line buffer 028 for prediction refers to the value as shown in FIG. 21, quantization results of a prediction device for an interlace image and a prediction device for a progressive image are represented as the tables shown in FIGS. 22 and 23. When the line buffer 028 for prediction has the pixel value as shown in FIG. 21, the tables shown in FIGS. 22 and 23 can be easily derived from the operation of the image data decompression (decoding) device according to the embodiments of the present invention shown in FIGS. 20A and 20B. In this case, the pixel C (029) after the previous decompression is the pixel B (030) after the current decompression, and the previous "prediction value + prediction error quantization value" is the current pixel A (027).

[0066] FIG. 24 is a block diagram of the configuration of the system of the image data decompression device according to the embodiment of the present invention, and is realized by a block represented as the function of the contents described above. In FIG. 24, the image data decompression device according to the embodiment of the present invention inputs the image data 202 as the output of the image data compression device to the decompressing unit 210, the decompressing unit 210 inputs the input image data 202to the inverse encoding unit 211 corresponding to the inverse encoder 022 shown in FIG. 14, and obtains a quantization number from the inverse encoding unit 211. Then, the quantization number is input to the inverse quantizing unit 212 corresponding to the inverse-quantizer 023. The inverse quantizing unit 212 obtains a prediction error quantization value from the quantization number with reference to the inverse quantization table 213 as shown in FIG. 15. The obtained prediction error quantization value is input to the predicting unit 214. The predicting unit 214 refers to the intermediate value correspondence table 215 as shown in FIG. 17, and the intermediate value - image format correspondence table 216 as shown in FIG. 18, calculates a candidate for a prediction value by the prediction value candidate calculation unit 217 corresponding to prediction value candidate calculation module 031, and determines a prediction value from among the candidates for a prediction value calculated by the prediction value candidate calculation unit 217 by the prediction value determination unit 218 corresponding to the MPX 035 on the basis of the image format 204. Then, the predicting unit 214 outputs the prediction error quantization value calculated by the prediction value candidate calculation unit 217 as output and a prediction value determined by the prediction value determination unit 218, and inputs the output prediction value and the prediction error quantization value to the decompression value calculation unit 219. The decompression value calculation unit 219 adds up the input prediction value and prediction error quantization value to obtain a decompressed pixel value, accumulates the value, and obtains the decompressed (image) data 206.

[0067] Thus, according to the image data decompression device of the present invention, the prediction value can be appropriately used depending on the image format, and the prediction error can be reduced on each image format, thereby enhancing the decompression efficiency. Since only a prediction value is to be switched depending on the image format, a program and a circuit configuration for implementation can be simplified.

[0068] In the descriptions above, an intermediate value among A, B, and C is obtained and a candidate for a prediction value to be selected is determined on the basis of an intermediate value correspondence table, but an intermediate value of candidates for a prediction value can also be obtained after calculating the candidates for a prediction value.

[0069] Also in the description above, the prediction value for an interlace image is (A+C)/2, A and (2A+C-B)/2. However, the prediction value candidate calculation module calculates "A", " (m*A+n*C) / (m+n) ", and "A+ (m*A+n*C) / (m+n) - (m*A+n*B) / (m+n) " using the coefficients m and n from the pixel level values A, B, and C, substitutes a value using coefficients m and n that are values other than 0, and can calculate a prediction value strong against the fluctuation of the pixel level value in the vertical direction using "A", "(A+3*C)/4", and "(4*A+3*C-3*B)/4" as the prediction value for an interlace image. Furthermore, the prediction value determination module and the prediction value calculation module can perform parallel processing, and a prediction value to be used can be determined after each candidate for a prediction value is calculated. Therefore, the parallel processing can perform data decompression at a high speed.

[0070] In addition, regardless of an interlace image or an progressive image as an image format, the prediction value can be used appropriately according to a signal indicating a sampling format of image data to attain similar effect.

[0071] For example, the sampling format of the image data stored on a common DVD is normally a 4:2:2 format. In this format, the image size of the Y component (brightness component) is 720 × 480, the horizontal image size of the C component (color component) is a half of the Y component. Therefore, since the C component refers to image data having a strong correlation in the vertical direction, problems can be solved by making the value of n larger than the value of m. That is, by increasing the gravity to the peripheral pixel having a strong correlation in the vertical direction (increasing the value of n), the degradation of the prediction error can avoided. For example, in the interlace image having the image data of a sampling format of 4:2:2 and the image size of 720 × 480, the image data having a strong correlation in the vertical direction is processed using m=1 and n=3 for the C component. For the Y component, image data having a strong correlation in the horizontal direction is processed using m=1 and n=1 in the case of the image size 720 × 480. Relating to the coefficients m and n when the image data has a strong correlation in the horizontal direction, the value of m is assigned the same value of n. If the degradation of the prediction error still advances, the value of m is further increased. Furthermore, when the sampling format of the image data as another example is a 4:4:4 format and the image size is 800 × 800 as a progressive image, m=0 and n=1.

[0072] Furthermore, since only field memory (buffer of 1 line) is used, and no frame memory is required, the implementation cost can reduced. Furthermore, since an interlace image and a progressive image can be processed only by switching prediction values, the circuit scale can be simplified.

[0073] In the description above, the prediction value candidate calculation module 031 calculates k (6 in the description above) candidates for a prediction value, and selects one prediction value by the multiplexer (MPX) 035 according to a control signal narrowed with reference to FIGS. 17 and 18. In the process order, a candidate for a prediction value can be narrowed with reference to FIGS. 17 and 18, and a prediction value can be calculated with the same timing, and the process speed can be improved. That is, in each of the prediction value determination module and the prediction value calculation module, parallel processing is performed and each candidate for a prediction value is calculated, each candidate for a prediction value is calculated, and then a prediction value to be used can be determined, thereby performing data compression at a high speed by the parallel processing. In the prediction value determination module and the prediction value calculation module, an arithmetic process is performed using a table. Therefore, program processing can be performed at a high speed.

[0074] On the other hand, as the implementation when the configuration scale is reduced without overweighting the process speed, the intermediate value correspondence table shown in FIG. 17 is removed from the configuration, three prediction values are first calculated depending on the image format with reference to FIG. 18, and a intermediate value is directly obtained from the three prediction values. Furthermore, another pattern can be calculating and determining by which equation of a candidate for a prediction value using the intermediate value correspondence table shown in FIG. 17, and performing a calculating process by the determined equation only, thereby deleting the configuration of selecting one prediction value from among a plurality of calculated prediction values.

Applicability in Industry

[0075] The image data compression device and the image data decompression device according to the present invention can be read as one-chip LSI to be loaded into a desired application equipment, and can be applied to, for example, a car-mounted video data transmission system including a car navigation system, various amusement (game, animation, etc.) equipment. For example, in a car-mounted video data transmission system, a display is provided on the rear side to supply video different from the front side. Therefore, the image data processed by the video data processing device implemented on the front side can be compressed and transmitted to the rear side. On the rear side, the transmitted image data is decompressed and displayed. When image data is relayed, the compressing and decompressing processes are repeated at each relay point.

Claims

1. An image data compression device (110), comprising

- a prediction device (112) for predicting a pixel level value of a pixel to be compressed when image data is compressed,

- a quantizing unit (117) for quantizing a difference value between a prediction value determined by the prediction device (112) and an actual pixel level value and associating a quantization number to the quantized difference,

- and a coding unit (119) performing encoding on the basis of the quantization number,

characterized in that the prediction device (112) includes:

- a prediction value candidate calculation unit (115) to calculate a plurality of candidates for a prediction value with respect to each of

-- an image format in which an image has a stronger pixel correlation in a horizontal direction than in a vertical direction and

-- an image format in which an image has a stronger pixel correlation in a vertical direction than in a horizontal direction,

using pixel level values of a pixel left to the pixel to be compressed defined as A, a pixel above the pixel to be compressed defined as C, and a pixel upper left to the pixel to be compressed defined as B; and

- a prediction value determination unit (116) configured

- to narrow the plurality of candidates for the prediction value into two candidates for the prediction value by selecting

-- the pixel whose pixel value is the second largest value among pixel level values of A, C, and B in a case where the pixel level values of A, C, and B differ from each other,

-- pixel B in a case where the pixel level value of B is less than a pixel level value of A and the pixel level value of B is equal to a pixel level value of C, and in a case where the pixel level value of B is larger than a pixel level value of C and the pixel level value of B is equal to a pixel level value of A,

-- pixel A in a case where the pixel level value of A is less than a pixel level value of C and the pixel level value of A is equal to a pixel level value of B, and in a case where the pixel level value of A is larger than a pixel level value of B and the pixel level value of A is equal to a pixel level value of C, and

-- pixel C in a case where the pixel level value of C is less than a pixel level value of B and the pixel level value of C is equal to a pixel level value of A, in a case where the pixel level value of C is larger than a pixel level value of A and the pixel level value of C is equal to a pixel level value of B, and in a case where the pixel level values of A, C, and B are same,

by associating via a first correspondence table an index value to the selected pixel, and selecting the two candidates corresponding to said index value via a second correspondence table;

- to determine one of the two candidates for the prediction value as the prediction value according to whether the image format of said image is

-- the image format in which an image has the stronger pixel correlation in a horizontal direction than in a vertical direction or

-- the image format in which an image has the stronger pixel correlation in a vertical direction than in a horizontal direction.

2. The device according to claim 1, wherein the image format in which an image has the stronger pixel correlation in the horizontal direction than in the vertical direction is an interlace image format, and the image format in which an image has the stronger pixel correlation in the vertical direction than in the horizontal direction is a progressive image format.

3. The device according to claim 1 or 2, wherein

- the prediction device further includes:

- - a multiplexer (015) to output the prediction value selected with a control signal corresponding to the candidate for the prediction value determined as the prediction value;

- - a plurality of registers to hold the peripheral pixels A, B and C; and

- - a line buffer (008) to hold one line of pixel values, and

- the prediction value determination unit (116) comprises means for transmitting the control signal to the multiplexer.

4. An image data compressing method by an image data compression device (110), comprising:

- a process of quantizing a difference value between a prediction value and an actual pixel level value and associating a quantization number to the quantized difference;

- a process of holding pixel values of a pixel A defined as a pixel left to the pixel to be compressed, a pixel C defined as a pixel above the pixel to be compressed, and a pixel B defined as a pixel upper left to the pixel to be compressed, before encoding data, and

- a process of encoding on the basis of the quantization number,
characterized by

- a process of calculating a plurality of candidates for the prediction value with respect to each of

-- an image format in which an image has a stronger pixel correlation in a horizontal direction than in a vertical direction and

-- an image format in which an image has a stronger pixel correlation in a vertical direction than in a horizontal direction,
using pixel level values of the held pixels A, C, and B;

- a process of

-- narrowing the plurality of candidates for the prediction value into two candidates for the prediction value by selecting

-- the pixel whose pixel value is the second largest value among pixel level values of A, C, and B in a case where the pixel level values of A, C, and B differ from each other,

-- pixel B in a case where the pixel level value of B is less than a pixel level value of A and the pixel level value of B is equal to a pixel level value of C, and in a case where the pixel level value of B is larger than a pixel level value of C and the pixel level value of B is equal to a pixel level value of A,

--- pixel A in a case where the pixel level value of A is less than a pixel level value of C and the pixel level value of A is equal to a pixel level value of B, and in a case where the pixel level value of A is larger than a pixel level value of B and the pixel level value of A is equal to a pixel level value of C, and

--- pixel C in a case where the pixel level value of C is less than a pixel level value of B and the pixel level value of C is equal to a pixel level value of A, in a case where the pixel level value of C is larger than a pixel level value of A and the pixel level value of C is equal to a pixel level value of B, and in a case where the pixel level values of A, C, and B are same
by associating via a first correspondence table an index value to the selected pixel, and selecting the two candidates corresponding to said index value via a second correspondence table;

-- determining one of the two candidates for the prediction value according to whether the image format of said image is

--- the image format in which an image has the stronger pixel correlation in a horizontal direction than in a vertical direction or

--- the image format in which an image has the stronger pixel correlation in a vertical direction than in a horizontal direction.

5. The method according to claim 4 , comprising

- a process of transmitting the a control signal corresponding to the candidate for the prediction value determined as the prediction value to a multiplexer, and

- a process of outputting the prediction value selected with the control signal by the multiplexer.

6. A program adapted to direct a computer to perform the method of claim 4.

7. The program according to claim 6, comprising

- a step of transmitting a control signal corresponding to the candidate for the prediction value determined as the prediction value to a multiplexer, and

- a step of outputting the prediction value selected with the control signal by the multiplexer.

8. An image data decompression device (210) for decompressing image data (202) comprising

- an inverse encoder (211) for receiving compressed code of image data and outputting a quantization number corresponding to the code,

- an inverse-quantizer (212) for obtaining from the quantization number a prediction error quantization value associated with the quantization number,

- a prediction device (214) for determining a prediction value, and

- a decompression value calculation unit (219) for adding the prediction value determined by the prediction device (214) to the prediction error quantization value obtained through the inverse-quantizer (212) thereby obtaining a decompressed pixel data;
characterized in that the prediction device (214) includes

- a prediction value candidate calculation unit (217) for

-- calculating a plurality of candidates for a prediction value with respect to each of

--- an image format in which an image has the stronger pixel correlation in a horizontal direction than in a vertical direction and

--- an image format in which an image has the stronger pixel correlation in a vertical direction than in a horizontal direction

using a pixel level values of a pixel left to the pixel to be decompressed defined A, a pixel above the pixel to be compressed defined C, and a pixel upper left to the pixel to be decompressed defined B; and

- a prediction value determination unit (218) for

-- narrowing the plurality of candidates for the prediction value into two candidates for the prediction value by selecting

--- the pixel whose pixel value is the second largest value among pixel level values of A, C, and B in a case where the pixel level values of A, C, and B differ from each other,

--- pixel B in a case where the pixel level value of B is less than a pixel level value of A and the pixel level value of B is equal to a pixel level value of C, and in a case where the pixel level value of B is larger than a pixel level value of C and the pixel level value of B is equal to a pixel level value of A,

--- pixel A in a case where the pixel level value of A is less than a pixel level value of C and the pixel level value of A is equal to a pixel level value of B, and in a case where the pixel level value of A is larger than a pixel level value of B and the pixel level value of A is equal to a pixel level value of C, and

--- pixel C in a case where the pixel level value of C is less than a pixel level value of B and the pixel level value of C is equal to a pixel level value of A, in a case where the pixel level value of C is larger than a pixel level value of A and the pixel level value of C is equal to a pixel level value of B, and in a case where the pixel level values of A, C, and B are same

by associating via a first correspondence table an index value to the selected pixel, and selecting the two candidates corresponding to said index value via a second correspondence table; and

-- determining one of the two prediction candidates for the prediction value as the prediction value according to whether the image format of said image is

--- the image format in which an image has the stronger pixel correlation in a horizontal direction than in a vertical direction or

--- the image format in which an image has the stronger pixel correlation in a vertical direction than in a horizontal direction.

9. An image data decompressing method by an image data decompression device (210) which decompresses image data comprising

- a process of receiving compressed code of image data and outputting a quantization number corresponding to the code,

- a process of obtaining from the quantization number a prediction error quantization value associated with the quantization number,

- a process of adding a prediction value to the obtained prediction error quantization value thereby obtaining a decompressed pixel data, and obtaining a next peripheral pixel value;

- a process of holding pixel values of a pixel A defined as a pixel left to the pixel to be decompressed, a pixel C defined as a pixel above the pixel to be decompressed, and a pixel B defined as a pixel upper left to the pixel to be decompressed, before decompressing a compressed code,
characterized by

- a process of calculating a plurality of candidates for the prediction value with respect to each of

-- an image format in which an image has a stronger pixel correlation in a horizontal direction than in a vertical direction and

-- an image format in which an image has a stronger pixel correlation in a vertical direction than in a horizontal direction
using the pixel values of the held peripheral pixels A, C, and B;

- a process of

-- narrowing the plurality of candidates for the prediction value into two candidates for the prediction value by selecting

--- the pixel whose pixel value is the second largest value among pixel level values of A, C, and B in a case where the pixel level values of A, C, and B differ from each other,

--- pixel B in a case where the pixel level value of B is less than a pixel level value of A and the pixel level value of B is equal to a pixel level value of C, and in a case where the pixel level value of B is larger than a pixel level value of C and the pixel level value of B is equal to a pixel level value of A,

--- pixel A in a case where the pixel level value of A is less than a pixel level value of C and the pixel level value of A is equal to a pixel level value of B, and in a case where the pixel level value of A is larger than a pixel level value of B and the pixel level value of A is equal to a pixel level value of C, and

--- pixel C in a case where the pixel level value of C is less than a pixel level value of B and the pixel level value of C is equal to a pixel level value of A, in a case where the pixel level value of C is larger than a pixel level value of A and the pixel level value of C is equal to a pixel level value of B, and in a case where the pixel level values of A, C, and B are same
by associating via a first correspondence table an index value to the selected pixel, and selecting the two candidates corresponding to said index value via a second correspondence table;,

-- determining as the prediction value one of the two candidates for the prediction value according to whether the image format of said image is

--- the image format in which an image has the stronger pixel correlation in a horizontal direction than in a vertical direction or

--- the image format in which an image has the stronger pixel correlation in a vertical direction than in a horizontal direction.

10. The method according to claim 9, comprising

- a process of transmitting a control signal corresponding to the candidate for the prediction value determined as the prediction value to a multiplexer, and

- a process of outputting the prediction value selected with the control signal by the multiplexer.

11. A program adapted to direct a computer to perform the method of claim 9.

12. The program according to claim 11 , comprising

- a step of transmitting a control signal corresponding to the candidate for the prediction value determined as the prediction value to a multiplexer, and

- a step of outputting the prediction value selected with the control signal by the multiplexer.

Ansprüche

1. Bilddatenkomprimierungsvorrichtung (110), umfassend:

- eine Vorhersagevorrichtung (112) zum Vorhersagen eines Pixelpegelwertes eines Pixels, das zu komprimieren ist, wenn Bilddaten komprimiert werden,

- eine Quantisierungseinheit (117) zum Quantisieren eines Differenzwertes zwischen einem durch die Vorhersagevorrichtung (112) bestimmten Vorhersagewert und einem tatsächlichen Pixelpegelwert und Zuordnen einer Quantisierungszahl zu der quantisierten Differenz

- und eine Codiereinheit (119) zum Ausführen einer Codierung auf der Basis der Quantisierungszahl,
dadurch gekennzeichnet, daß die Vorhersagevorrichtung (112) enthält:

- eine Vorhersagewertkandidatenberechnungseinheit (115), um eine Vielzahl von Kandidaten für einen Vorhersagewert zu berechnen, in Bezug auf

- ein Bildformat, in dem ein Bild eine stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, als auch

- ein Bildformat, in dem ein Bild eine stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat,
unter Verwendung von Pixelpegelwerten eines Pixels links von dem zu komprimierenden Pixel, das als A definiert ist, eines Pixels oberhalb des zu komprimierenden Pixels, das als C definiert ist, und eines Pixels oben links von dem zu komprimierenden Pixel, das als B definiert ist; und

- eine Vorhersagewertbestimmungseinheit (116), die konfiguriert ist,

- um die Vielzahl von Kandidaten für den Vorhersagewert auf zwei Kandidaten für den Vorhersagewert einzuschränken, durch Selektieren

-- des Pixels, dessen Pixelwert der zweitgrößte Wert unter Pixelpegelwerten von A, C und B ist, falls sich die Pixelpegelwerte von A, C und B voneinander unterscheiden,

-- des Pixels B, falls der Pixelpegelwert von B kleiner als ein Pixelpegelwert von A ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von C ist und falls der Pixelpegelwert von B größer als ein Pixelpegelwert von C ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von A ist,

-- des Pixels A, falls der Pixelpegelwert von A kleiner als ein Pixelpegelwert von C ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von B ist und falls der Pixelpegelwert von A größer als ein Pixelpegelwert von B ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von C ist, und

-- des Pixels C, falls der Pixelpegelwert von C kleiner als ein Pixelpegelwert von B ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von A ist, falls der Pixelpegelwert von C größer als ein Pixelpegelwert von A ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von B ist und falls die Pixelpegelwerte von A, C und B dieselben sind,
durch Zuordnen, über eine erste Entsprechungstabelle, eines Indexwertes zu dem selektierten Pixel und Selektieren der zwei Kandidaten entsprechend dem genannten Indexwert über eine zweite Entsprechungstabelle;

- um einen der zwei Kandidaten für den Vorhersagewert als Vorhersagewert danach zu bestimmen, ob das Bildformat des genannten Bildes

-- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, oder

-- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat.

2. Vorrichtung nach Anspruch 1, bei der das Bildformat, in dem ein Bild die stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, ein Interlace-Bildformat ist und das Bildformat, in dem ein Bild die stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat, ein Progressive-Bildformat ist.

3. Vorrichtung nach Anspruch 1 oder 2, bei der

- die Vorhersagevorrichtung ferner enthält:

- einen Multiplexer (015), um den mit einem Steuersignal selektierten Vorhersagewert entsprechend dem Kandidaten für den Vorhersagewert auszugeben, der als Vorhersagewert bestimmt wurde;

- - eine Vielzahl von Registern, um die peripheren Pixel A, B und C zu halten; und

- - einen Zeilenpuffer (008), um eine Zeile von Pixelwerten zu halten, und

- die Vorhersagewertbestimmungseinheit (116) ein Mittel zum Senden des Steuersignals an den Multiplexer umfaßt.

4. Bilddatenkomprimierungsverfahren durch eine Bilddatenkomprimierungsvorrichtung (110), umfassend:

- einen Prozeß zum Quantisieren eines Differenzwertes zwischen einem Vorhersagewert und einem tatsächlichen Pixelpegelwert und Zuordnen einer Quantisierungszahl zu der quantisierten Differenz;

- einen Prozeß zum Halten von Pixelwerten eines Pixels A, das als Pixel links von dem zu komprimierenden Pixel definiert ist, eines Pixels C, das als Pixel oberhalb des zu komprimierenden Pixels definiert ist, und eines Pixels B, das als Pixel oben links von dem zu komprimierenden Pixel definiert ist, vor dem Codieren von Daten und

- einen Prozeß zum Codieren auf der Basis der Quantisierungszahl,
gekennzeichnet durch

- einen Prozeß zum Berechnen einer Vielzahl von Kandidaten für den Vorhersagewert in Bezug auf

-- ein Bildformat, in dem ein Bild eine stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, als auch

-- ein Bildformat, in dem ein Bild eine stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat,

unter Verwendung von Pixelpegelwerten der gehaltenen Pixel A, C und B;

- einen Prozeß zum

-- Einschränken der Vielzahl von Kandidaten für den Vorhersagewert auf zwei Kandidaten für den Vorhersagewert durch Selektieren

--- des Pixels, dessen Pixelwert der zweitgrößte Wert unter Pixelpegelwerten von A, C und B ist, falls sich die Pixelpegelwerte von A, C und B voneinander unterscheiden,

--- des Pixels B, falls der Pixelpegelwert von B kleiner als ein Pixelpegelwert von A ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von C ist und falls der Pixelpegelwert von B größer als ein Pixelpegelwert von C ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von A ist,

--- des Pixels A, falls der Pixelpegelwert von A kleiner als ein Pixelpegelwert von C ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von B ist und falls der Pixelpegelwert von A größer als ein Pixelpegelwert von B ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von C ist, und

--- des Pixels C, falls der Pixelpegelwert von C kleiner als ein Pixelpegelwert von B ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von A ist, falls der Pixelpegelwert von C größer als ein Pixelpegelwert von A ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von B ist und falls die Pixelpegelwerte von A, C und B dieselben sind,

durch Zuordnen, über eine erste Entsprechungstabelle, eines Indexwertes zu dem selektierten Pixel und Selektieren der zwei Kandidaten entsprechend dem genannten Indexwert über eine zweite Entsprechungstabelle;

-- Bestimmen eines der zwei Kandidaten für den Vorhersagewert danach, ob das Bildformat des genannten Bildes

--- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, oder

--- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat.

5. Verfahren nach Anspruch 4, umfassend:

- einen Prozeß zum Senden eines Steuersignals entsprechend dem Kandidaten für den Vorhersagewert, der als Vorhersagewert bestimmt wurde, an einen Multiplexer und

- einen Prozeß zum Ausgeben des mit dem Steuersignal selektierten Vorhersagewertes durch den Multiplexer.

6. Programm, das dafür ausgelegt ist, um einen Computer anzuweisen, das Verfahren nach Anspruch 4 auszuführen.

7. Programm nach Anspruch 6, umfassend:

- einen Schritt zum Senden eines Steuersignals entsprechend dem Kandidaten für den Vorhersagewert, der als Vorhersagewert bestimmt wurde, an einen Multiplexer und

- einen Schritt zum Ausgeben des mit dem Steuersignal selektierten Vorhersagewertes durch den Multiplexer.

8. Bilddatendekomprimierungsvorrichtung (210) zum Dekomprimieren von Bilddaten (202), umfassend:

- einen inversen Codierer (211) zum Empfangen eines komprimierten Codes von Bilddaten und Ausgeben einer Quantisierungszahl entsprechend dem Code,

- einen inversen Quantisierer (212) zum Erhalten, von der Quantisierungszahl, eines Vorhersagefehlerquantisierungswertes, der der Quantisierungszahl zugeordnet ist,

- eine Vorhersagevorrichtung (214) zum Bestimmen eines Vorhersagewertes und

- eine Dekomprimierungswertberechnungseinheit (219) zum Addieren des durch die Vorhersagevorrichtung (214) bestimmten Vorhersagewertes zu dem Vorhersagefehlerquantisierungswert, der durch den inversen Quantisierer (212) erhalten wurde, wodurch dekomprimierte Pixeldaten erhalten werden;
dadurch gekennzeichnet, daß die Vorhersagevorrichtung (214) enthält:

- eine Vorhersagewertkandidatenberechnungseinheit (217) zum

-- Berechnen einer Vielzahl von Kandidaten für einen Vorhersagewert in Bezug auf

--- ein Bildformat, in dem ein Bild die stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, als auch

--- ein Bildformat, in dem ein Bild die stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat,

unter Verwendung von Pixelpegelwerten eines Pixels links von dem zu dekomprimierenden Pixel, das als A definiert ist, eines Pixels oberhalb des zu dekomprimierenden Pixels, das als C definiert ist, und eines Pixels oben links von dem zu dekomprimierenden Pixel, das als B definiert ist; und

- eine Vorhersagewertbestimmungseinheit (218) zum

-- Einschränken der Vielzahl von Kandidaten für den Vorhersagewert auf zwei Kandidaten für den Vorhersagewert durch Selektieren

--- des Pixels, dessen Pixelwert der zweitgrößte Wert unter Pixelpegelwerten von A, C und B ist, falls sich die Pixelpegelwerte von A, C und B voneinander unterscheiden,

--- des Pixels B, falls der Pixelpegelwert von B kleiner als ein Pixelpegelwert von A ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von C ist und falls der Pixelpegelwert von B größer als ein Pixelpegelwert von C ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von A ist,

--- des Pixels A, falls der Pixelpegelwert von A kleiner als ein Pixelpegelwert von C ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von B ist und falls der Pixelpegelwert von A größer als ein Pixelpegelwert von B ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von C ist, und

--- des Pixels C, falls der Pixelpegelwert von C kleiner als ein Pixelpegelwert von B ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von A ist, falls der Pixelpegelwert von C größer als ein Pixelpegelwert von A ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von B ist und falls die Pixelpegelwerte von A, C und B dieselben sind,

durch Zuordnen, über eine erste Entsprechungstabelle, eines Indexwertes zu dem selektierten Pixel und Selektieren der zwei Kandidaten entsprechend dem genannten Indexwert über eine zweite Entsprechungstabelle; und

-- Bestimmen eines der zwei Vorhersagekandidaten für den Vorhersagewert als Vorhersagewert danach, ob das Bildformat des genannten Bildes

--- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, oder

--- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat.

9. Bilddatendekomprimierungsverfahren durch eine Bilddatendekomprimierungsvorrichtung (210), die Bilddaten dekomprimiert, umfassend:

- einen Prozeß zum Empfangen eines komprimierten Codes von Bilddaten und Ausgeben einer Quantisierungszahl entsprechend dem Code,

- einen Prozeß zum Erhalten, von der Quantisierungszahl, eines Vorhersagefehlerquantisierungswertes, der der Quantisierungszahl zugeordnet ist,

- einen Prozeß zum Addieren eines Vorhersagewertes zu dem erhaltenen Vorhersagefehlerquantisierungswert, wodurch dekomprimierte Pixeldaten erhalten werden, und Erhalten eines nächsten peripheren Pixelwertes;

- einen Prozeß zum Halten von Pixelwerten eines Pixels A, das als Pixel links von dem zu dekomprimierenden Pixel definiert ist, eines Pixels C, das als Pixel oberhalb des zu dekomprimierenden Pixels definiert ist, und eines Pixels B, das als Pixel oben links von dem zu dekomprimierenden Pixel definiert ist, vor dem Dekomprimieren eines komprimierten Codes,
gekennzeichnet durch

- einen Prozeß zum Berechnen einer Vielzahl von Kandidaten für den Vorhersagewert in Bezug auf

-- ein Bildformat, in dem ein Bild eine stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, als auch

-- ein Bildformat, in dem ein Bild eine stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat,

unter Verwendung der Pixelpegelwerte der gehaltenen peripheren Pixel A, C und B;

- einen Prozeß zum

-- Einschränken der Vielzahl von Kandidaten für den Vorhersagewert auf zwei Kandidaten für den Vorhersagewert durch Selektieren

--- des Pixels, dessen Pixelwert der zweitgrößte Wert unter Pixelpegelwerten von A, C und B ist, falls sich die Pixelpegelwerte von A, C und B voneinander unterscheiden,

--- des Pixels B, falls der Pixelpegelwert von B kleiner als ein Pixelpegelwert von A ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von C ist und falls der Pixelpegelwert von B größer als ein Pixelpegelwert von C ist und der Pixelpegelwert von B gleich einem Pixelpegelwert von A ist,

--- des Pixels A, falls der Pixelpegelwert von A kleiner als ein Pixelpegelwert von C ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von B ist und falls der Pixelpegelwert von A größer als ein Pixelpegelwert von B ist und der Pixelpegelwert von A gleich einem Pixelpegelwert von C ist, und

--- des Pixels C, falls der Pixelpegelwert von C kleiner als ein Pixelpegelwert von B ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von A ist, falls der Pixelpegelwert von C größer als ein Pixelpegelwert von A ist und der Pixelpegelwert von C gleich einem Pixelpegelwert von B ist und falls die Pixelpegelwerte von A, C und B dieselben sind,
durch Zuordnen, über eine erste Entsprechungstabelle, eines Indexwertes zu dem selektierten Pixel und Selektieren der zwei Kandidaten entsprechend dem genannten Indexwert über eine zweite Entsprechungstabelle;

-- Bestimmen, als Vorhersagewert, eines der zwei Kandidaten danach, ob das Bildformat des genannten Bildes

--- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in horizontaler Richtung als in vertikaler Richtung hat, oder

--- das Bildformat ist, in dem ein Bild die stärkere Pixelkorrelation in vertikaler Richtung als in horizontaler Richtung hat.

10. Verfahren nach Anspruch 9, umfassend:

- einen Prozeß zum Senden eines Steuersignals entsprechend dem Kandidaten für den Vorhersagewert, der als Vorhersagewert bestimmt wurde, an einen Multiplexer und

- einen Prozeß zum Ausgeben des mit dem Steuersignal selektierten Vorhersagewertes durch den Multiplexer.

11. Programm, das dafür ausgelegt ist, einen Computer anzuweisen, das Verfahren nach Anspruch 9 auszuführen.

12. Programm nach Anspruch 11, umfassend:

- einen Schritt zum Senden eines Steuersignals entsprechend dem Kandidaten für den Vorhersagewert, der als Vorhersagewert bestimmt wurde, an einen Multiplexer und

- einen Schritt zum Ausgeben des mit dem Steuersignal selektierten Vorhersagewertes durch den Multiplexer.

Revendications

1. Dispositif de compression de données d'image (110), comprenant :

- un dispositif de prédiction (112) pour prédire une valeur de niveau de pixel d'un pixel à compresser lorsque des données d'image sont compressées,

- une unité de quantification (117) pour quantifier une valeur de différence entre une valeur de prédiction déterminée par le dispositif de prédiction (112) et une valeur de niveau réel de pixel et associer un nombre de quantification à la différence quantifiée,

- et une unité de codage (119) effectuant un codage en se basant sur le nombre de quantification,
caractérisé en ce que le dispositif de prédiction (112) comporte :

- une unité de calcul de candidats de valeur de prédiction (115) pour calculer une pluralité de candidats pour une valeur de prédiction concernant chaque format parmi

-- un format d'image dans lequel une image présente une corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, et

-- un format d'image dans lequel une image présente une corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale,
en utilisant des valeurs de niveau de pixel d'un pixel à gauche du pixel à compresser défini par A, un pixel au-dessus du pixel à compresser défini par C, et un pixel en haut et à gauche du pixel à compresser défini par B ; et

- une unité de détermination de valeur de prédiction (116) configurée

- pour restreindre la pluralité de candidats pour la valeur de prédiction à deux candidats pour la valeur de prédiction en sélectionnant

-- le pixel dont la valeur de pixel est la deuxième valeur la plus grande parmi les valeurs de niveau de pixel de A, C et B dans le cas où les valeurs de niveau de pixel de A, C et B diffèrent entre elles,

-- le pixel B dans le cas où la valeur de niveau de pixel de B est inférieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de C, et dans le cas où la valeur de niveau de pixel de B est supérieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de A,

-- le pixel A dans le cas où la valeur de niveau de pixel de A est inférieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de B, et dans le cas où la valeur de niveau de pixel de A est supérieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de C, et

-- le pixel C dans le cas où la valeur de niveau de pixel de C est inférieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de A, dans le cas où la valeur de niveau de pixel de C est supérieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de B, et dans le cas où les valeurs de niveau de pixel de A, C et B sont les mêmes,
en associant par l'intermédiaire d'un premier tableau de correspondance une valeur d'index au pixel sélectionné et en sélectionnant les deux candidats correspondants à ladite valeur d'index par l'intermédiaire d'un second tableau de correspondance ;

- pour déterminer l'un des deux candidats pour la valeur de prédiction en tant que valeur de prédiction selon que le format d'image de ladite image est

-- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, ou

-- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale.

2. Dispositif selon la revendication 1, dans lequel le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale est un format d'image entrelacé et le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale est un format d'image progressif.

3. Dispositif selon la revendication 1 ou 2, dans lequel

- le dispositif de prédiction comporte en outre :

-- un multiplexeur (015) pour fournir en sortie la valeur de prédiction sélectionnée avec un signal de commande correspondant au candidat pour la valeur de prédiction déterminée en tant que valeur de prédiction ;

-- une pluralité de registres pour contenir les pixels périphériques A, B et C;et

-- un tampon de ligne (008) pour contenir une ligne de valeurs de pixel, et

- l'unité de détermination de valeur de prédiction (116) comprend un moyen pour transmettre le signal de commande au multiplexeur.

4. Procédé de compression de données d'image au moyen d'un dispositif de compression de données d'image (110), comprenant

- un processus de quantification d'une valeur de différence entre une valeur de prédiction et une valeur de niveau réel de pixel et d'association d'un nombre de quantification à la différence quantifiée ;

- un processus de maintien de valeurs de pixel d'un pixel A défini comme un pixel à gauche du pixel à compresser, un pixel C défini comme un pixel au-dessus du pixel à compresser, et un pixel B défini comme un pixel en haut et à gauche du pixel à compresser, avant de coder les données, et

- un processus de codage se basant sur le nombre de quantification, caractérisé par

- un processus de calcul d'une pluralité de candidats pour la valeur de prédiction concernant chaque format parmi

-- un format d'image dans lequel une image présente une corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, et

-- un format d'image dans lequel une image présente une corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale,
l'utilisation des valeurs de niveau de pixel des pixels maintenus A, C et B ;

- un processus

-- de restriction de la pluralité de candidats pour la valeur de prédiction à deux candidats pour la valeur de prédiction en sélectionnant

--- le pixel dont la valeur de pixel est la deuxième valeur la plus grande parmi les valeurs de niveau de pixel de A, C et B dans le cas où les valeurs de niveau de pixel de A, C et B diffèrent entre elles,

--- le pixel B dans le cas où la valeur de niveau de pixel de B est inférieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de C, et dans le cas où la valeur de niveau de pixel de B est supérieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de A,

--- le pixel A dans le cas où la valeur de niveau de pixel de A est inférieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de B, et dans le cas où la valeur de niveau de pixel de A est supérieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de C, et

--- le pixel C dans le cas où la valeur de niveau de pixel de C est inférieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de A, dans le cas où la valeur de niveau de pixel de C est supérieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de B, et dans le cas où les valeurs de niveau de pixel de A, C et B sont les mêmes,
en associant par l'intermédiaire d'un premier tableau de correspondance une valeur d'index au pixel sélectionné et en sélectionnant les deux candidats correspondants à ladite valeur d'index par l'intermédiaire d'un second tableau de correspondance ;

-- de détermination de l'un des deux candidats pour la valeur de prédiction selon que le format d'image de ladite image est

--- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, ou

--- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale.

5. Procédé selon la revendication 4, comprenant

- un processus de transmission à un multiplexeur d'un signal de commande correspondant au candidat pour la valeur de prédiction déterminée en tant que valeur de prédiction, et

- un processus de fourniture en sortie de la valeur de prédiction sélectionnée avec le signal de commande par le multiplexeur.

6. Programme adapté à faire exécuter par un ordinateur le procédé selon la revendication 4.

7. Programme selon la revendication 6 comprenant

- une étape de transmission à un multiplexeur d'un signal de commande correspondant au candidat pour la valeur de prédiction déterminée en tant que valeur de prédiction, et

- une étape de fourniture en sortie de la valeur de prédiction sélectionnée avec le signal de commande par le multiplexeur.

8. Dispositif de décompression de données d'images (210) pour décompresser des données d'images (202) comprenant

- un codeur inverse (211) pour recevoir un code de données d'image compressées et fournir en sortie un nombre de quantification correspondant au code,

- un quantificateur inverse (212) pour obtenir d'après le nombre de quantification une valeur de quantification d'erreur de prédiction associée au nombre de quantification,

- un dispositif de prédiction (214) pour déterminer une valeur de prédiction, et

- une unité de calcul de valeur de décompression (219) pour ajouter la valeur de prédiction déterminée par le dispositif de prédiction (214) à la valeur de quantification d'erreur de prédiction obtenue par l'intermédiaire du quantificateur inverse (212) de façon à obtenir des données de pixel décompressées ;
caractérisé en ce que le dispositif de prédiction (214) comporte

- une unité de calcul de candidats de valeur de prédiction (217) pour

-- calculer une pluralité de candidats pour une valeur de prédiction concernant chaque format parmi

--- un format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, et

--- un format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale,
en utilisant des valeurs de niveau de pixel d'un pixel à gauche du pixel à compresser défini par A, un pixel au-dessus du pixel à décompresser défini par C, et un pixel en haut et à gauche du pixel à décompresser défini par B ; et

- une unité de détermination de valeur de prédiction (218) pour

-- restreindre la pluralité de candidats pour la valeur de prédiction à deux candidats pour la valeur de prédiction en sélectionnant

--- le pixel dont la valeur de pixel est la deuxième valeur la plus grande parmi les valeurs de niveau de pixel de A, C et B dans le cas où les valeurs de niveau de pixel de A, C et B diffèrent entre elles,

--- le pixel B dans le cas où la valeur de niveau de pixel de B est inférieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de C, et dans le cas où la valeur de niveau de pixel de B est supérieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de A,

--- le pixel A dans le cas où la valeur de niveau de pixel de A est inférieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de B, et dans le cas où la valeur de niveau de pixel de A est supérieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de C, et

--- le pixel C dans le cas où la valeur de niveau de pixel de C est inférieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de A, dans le cas où la valeur de niveau de pixel de C est supérieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de B, et dans le cas où les valeurs de niveau de pixel de A, C et B sont les mêmes,
en associant par l'intermédiaire d'un premier tableau de correspondance une valeur d'index au pixel sélectionné et en sélectionnant les deux candidats correspondants à ladite valeur d'index par l'intermédiaire d'un second tableau de correspondance ; et

-- déterminer l'un des deux candidats pour la valeur de prédiction en tant que valeur de prédiction selon que le format d'image de ladite image est

--- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, ou

--- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale.

9. Procédé de décompression de données d'image au moyen d'un dispositif de décompression de données d'image (210) qui décompresse des données d'image, comprenant :

- un processus de réception d'un code compressé de données d'image et de fourniture en sortie d'un nombre de quantification correspondant au code,

- un processus d'obtention d'après le nombre de quantification d'une valeur de quantification d'erreur de prédiction associée au nombre de quantification,

- un processus d'addition d'une valeur de prédiction à la valeur de quantification d'erreur de prédiction obtenue de façon à obtenir des données de pixel décompressées et à obtenir une valeur de pixel périphérique suivant ;

- un processus de maintien de valeurs de pixel d'un pixel A défini comme un pixel à gauche du pixel à décompresser, un pixel C défini comme un pixel au-dessus du pixel à décompresser, et un pixel B défini comme un pixel en haut et à gauche du pixel à décompresser, avant de décompresser un code compressé,
caractérisé par

- un processus de calcul d'une pluralité de candidats pour une valeur de prédiction concernant chaque format parmi

-- un format d'image dans lequel une image présente une corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, et

-- un format d'image dans lequel une image présente une corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale,
en utilisant les valeurs de pixel des pixels maintenus A, C et B ;

- un processus de

-- restriction de la pluralité de candidats pour la valeur de prédiction à deux candidats pour la valeur de prédiction en sélectionnant

--- le pixel dont la valeur de pixel est la deuxième valeur la plus grande parmi les valeurs de niveau de pixel de A, C et B dans le cas où les valeurs de niveau de pixel de A, C et B diffèrent entre elles,

--- le pixel B dans le cas où la valeur de niveau de pixel de B est inférieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de C, et dans le cas où la valeur de niveau de pixel de B est supérieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de B est égale à la valeur de niveau de pixel de A,

--- le pixel A dans le cas où la valeur de niveau de pixel de A est inférieure à la valeur de niveau de pixel de C et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de B, et dans le cas où la valeur de niveau de pixel de A est supérieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de A est égale à la valeur de niveau de pixel de C, et

--- le pixel C dans le cas où la valeur de niveau de pixel de C est inférieure à la valeur de niveau de pixel de B et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de A, dans le cas où la valeur de niveau de pixel de C est supérieure à la valeur de niveau de pixel de A et la valeur de niveau de pixel de C est égale à la valeur de niveau de pixel de B, et dans le cas où les valeurs de niveau de pixel de A, C et B sont les mêmes,
en associant par l'intermédiaire d'un premier tableau de correspondance une valeur d'index au pixel sélectionné et en sélectionnant les deux candidats correspondants à ladite valeur d'index par l'intermédiaire d'un second tableau de correspondance ;

-- détermination en tant que valeur de prédiction de l'un des deux candidats pour la valeur de prédiction selon que le format d'image de ladite image est

--- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction horizontale que dans la direction verticale, ou

--- le format d'image dans lequel une image présente la corrélation de pixel plus forte dans la direction verticale que dans la direction horizontale.

10. Procédé selon la revendication 9, comprenant

- un processus de transmission à un multiplexeur d'un signal de commande correspondant au candidat pour la valeur de prédiction déterminée en tant que valeur de prédiction, et

- un processus de fourniture en sortie de la valeur de prédiction sélectionnée avec le signal de commande par le multiplexeur.

11. Programme adapté à faire exécuter par un ordinateur le procédé selon la revendication 9.

12. Programme selon la revendication 11, comprenant

- une étape de transmission à un multiplexeur d'un signal de commande correspondant au candidat pour la valeur de prédiction déterminée en tant que valeur de prédiction, et

- une étape de fourniture en sortie de la valeur de prédiction sélectionnée avec le signal de commande par le multiplexeur.

Drawing